Advancing Reservoir Characterization in Hassi Messaoud: Dynamic Well Test Analysis with Physics-Constrained LSTM Networks.

Saved in:

Bibliographic Details
Title:	Advancing Reservoir Characterization in Hassi Messaoud: Dynamic Well Test Analysis with Physics-Constrained LSTM Networks.
Authors:	Hadjadj, Asma¹ as.hadjadj@lagh-univ.dz, Bouarar, Fahima¹, Youcefi, Mohammed Riad², Abdellah Taouti, Mohamed Ben¹
Source:	Petroleum & Coal. 2026, Vol. 68 Issue 1, p42-50. 9p.
Subject Terms:	Long short-term memory, Dynamic testing, Oil fields, Permeability measurement, Petroleum reservoirs, Artificial intelligence
Geographic Terms:	Algeria
Abstract:	Reservoir characterization remains one of the most challenging yet crucial aspects of petroleum engineering, directly impacting production optimization and field development decisions. Traditional methods for predicting reservoir permeability often struggle with complex geological conditions and interpretative challenges. In this research, we present an innovative solution that utilizes Long ShortTerm Memory (LSTM) neural networks to predict permeability from dynamic well test data, fundamentally enhancing the approach to reservoir analysis. Our methodology was applied to real field data from two production wells in Algeria’s Hassi Messaoud field, using time-series measurements of pressure, flow rate, and time as model inputs. The LSTM-based approach significantly outperformed conventional well test analysis, offering improved accuracy in permeability prediction and demonstrating notable resilience to complex reservoir behaviors and data noise. The findings indicate that this machine learning approach automates and streamlines interpretation, reducing dependence on specialized expertise and making advanced reservoir characterization more accessible. This methodology holds substantial promise for real-time reservoir management, providing petroleum engineers with a robust tool to make informed decisions in challenging geological environments. This work highlights the transformative potential of integrating artificial intelligence into traditional reservoir engineering practices, paving the way for more efficient, accurate, and reliable reservoir characterization throughout the industry. [ABSTRACT FROM AUTHOR]
Database:	Energy & Power Source

Description
Abstract:	Reservoir characterization remains one of the most challenging yet crucial aspects of petroleum engineering, directly impacting production optimization and field development decisions. Traditional methods for predicting reservoir permeability often struggle with complex geological conditions and interpretative challenges. In this research, we present an innovative solution that utilizes Long ShortTerm Memory (LSTM) neural networks to predict permeability from dynamic well test data, fundamentally enhancing the approach to reservoir analysis. Our methodology was applied to real field data from two production wells in Algeria’s Hassi Messaoud field, using time-series measurements of pressure, flow rate, and time as model inputs. The LSTM-based approach significantly outperformed conventional well test analysis, offering improved accuracy in permeability prediction and demonstrating notable resilience to complex reservoir behaviors and data noise. The findings indicate that this machine learning approach automates and streamlines interpretation, reducing dependence on specialized expertise and making advanced reservoir characterization more accessible. This methodology holds substantial promise for real-time reservoir management, providing petroleum engineers with a robust tool to make informed decisions in challenging geological environments. This work highlights the transformative potential of integrating artificial intelligence into traditional reservoir engineering practices, paving the way for more efficient, accurate, and reliable reservoir characterization throughout the industry. [ABSTRACT FROM AUTHOR]
ISSN:	13353055